不平衡数据集下特征词两面性的新型降维算法  

Novel feature selection method based on two-side considering imbalance problem

作　　者：付鑫[1] 王洪国[1] 邵增珍[1] 杜秋霞[1] Fu Xin;Wang Hongguo;Shao Zengzhen;Du Qiuxia(Shandong Provincial Logistics Optimization ＆ Predictive Engineering Technology Research Center,School of Information Science ＆ Enginee-ring,Shandong Normal University,Jinan 250014,China)

机构地区：[1]山东师范大学信息科学与工程学院山东省物流优化与预测工程技术研究中心,济南250014

出　　处：《计算机应用研究》2018年第7期1947-1949,1969,共4页Application Research of Computers

基　　金：山东省科技发展计划资助项目(2014GGH201022);山东省经信委软科学计划资助项目(2015EI010);国家自然科学基金资助项目(71461027)

摘　　要：传统DFS特征选择算法在降维处理时既未考虑样本分布不均的情况,又未涉及负特征词对类别的影响。综合考虑DFS的缺陷并进行优化处理,将DFS与卡方检测算法CHI结合,提出一种改进型特征选择算法DFS-sCHI。引入负特证词作为类别划分的影响因子之一,解决不平衡数据集下所提特征词类别分布不均的问题。经实验分析,不平衡数据集下,DFS-sCHI相比较于DFS在分类精度上有明显提高。DFS（ distinguishing feature selector） neither considered the situation of uneven distribution of samples nor involved the impact of negative words for category. This paper considered the defects of DFS and optimized these defects,combined DFS with CHI to proposed DFS-s CHI that was a improve feature selection method. This method introduced the power of negative words in order to solve the problem that the feature of category was uneven selection. The experimental results indicate that the proposed method improves the classification accuracy obviously.

关 键 词：不平衡数据集 文本分类 特征选择 DFS-sCHI 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]